Development manual for vsearch 2.32.0, built from the dev branch: it describes changes that are not released yet. The manual for the current release is published separately.

NAME

vsearch --fastq_convert — convert between fastq format variants

SYNOPSIS

vsearch --fastq_convert fastqfile --fastqout filename [options]

DESCRIPTION

The vsearch command --fastq_convert converts a fastq file from one quality-score encoding to another. The input encoding is specified with --fastq_ascii (default: 33) and the output encoding with --fastq_asciiout (default: 33). Both accept the values 33 (phred+33, Sanger/Illumina 1.8+) and 64 (phred+64, Illumina 1.3+/1.5+). The older Solexa/Illumina 1.0 format shares the offset 64 but defines its scores differently (see vsearch-fastq(5)); --fastq_solexa converts such a file to the Phred scale, and is the only place in vsearch where the Solexa score definition is understood.

Quality scores are remapped during conversion. Output scores are always clamped to the range set by --fastq_qminout and --fastq_qmaxout, whose defaults are 0 and 93: scores are written unchanged unless --fastq_qmaxout is lowered (before version 2.32.0 the default was 41, and any higher score was silently reduced to it), and negative scores (e.g. Solexa scores read with --fastq_ascii 64 --fastq_qmin -5) are raised to 0, unless the bounds are changed. The input score range is validated against --fastq_qmin and --fastq_qmax.

Use --fastq_chars (see vsearch-fastq_chars(1)) to detect the encoding of an unknown fastq file before converting.

To illustrate a conversion from phred+64 to phred+33:

Input (phred+64):    Output (--fastq_ascii 64 --fastq_asciiout 33):

@s1                  @s1
ACGT                 ACGT
+                    +
hijk                 IJKL    (same Phred scores, different ASCII offset)

Converting Solexa scores

A Solexa score is Q = -10 log10(p / (1 - p)), while every score vsearch understands is a Phred score, Q = -10 log10(p). The two share the ASCII offset 64 and nothing else, so rebasing the offset alone leaves the scores wrong. --fastq_solexa applies

Q_phred = 10 log10(10^(Q_solexa / 10) + 1)

and rounds the result to the nearest integer. It requires --fastq_ascii 64, and implies --fastq_qmin -5 (the floor of the Solexa scale) unless --fastq_qmin is given explicitly.

The mapping is the identity from Solexa 10 upward, so only the fifteen lowest scores change — which are exactly the ones a quality filter acts on:

Solexa:  -5  -4  -3  -2  -1   0   1   2   3   4   5   6   7   8   9  10
Phred:    1   1   2   2   3   3   4   4   5   5   6   7   8   9  10  10

The conversion is one-way and lossy: six Solexa pairs collapse onto a single Phred score each ({-5, -4}, {-3, -2}, {-1, 0}, {1, 2}, {3, 4} and {9, 10}), so the original scores cannot be recovered afterwards. There is no option to write Solexa scores. This is inherent to the two scales rather than to the implementation.

OPTIONS

mandatory options

--fastqout filename: Write converted sequences to filename, in fastq format (see vsearch-fastq(5)).

core options

--fastq_ascii 33|64: Specify the offset used as the basis for the fastq quality score when reading fastq files. For example, an offset of 33 means that a quality value of 41 is represented by the 74th ASCII symbol (33 + 41 = 74), which is ‘J’. See ascii(7) for a view of the ASCII character set. The offset value is either 33 or 64, default is 33.

The offset matters even to a command that never decodes a quality score: it is what the reader compares the observed quality symbols against before warning that the file may use the other encoding, it sets the default of --fastq_qmax (the highest score the offset can represent), and the sum rules on --fastq_qmin and --fastq_qmax are stated in terms of it.

--fastq_asciiout 33|64: Specify the offset used as the basis for the fastq quality score when writing fastq output files. For example, an offset of 33 means that a quality value of 41 is represented by the 74th ASCII symbol (33 + 41 = 74), which is ‘J’. See ascii(7) for a view of the ASCII character set. The offset value is either 33 or 64, default is 33.
--fastq_qmaxout integer: Specify the maximum quality score used when writing fastq files. The default is the highest score the output offset can represent (93 with --fastq_asciiout 33, 62 with offset 64), so scores read from the input are written back unchanged. Before version 2.32.0 the default was 41, the usual maximum for Sanger/Illumina 1.8+ files, which silently reduced any higher score. Older formats may use a maximum quality score of 40. Two commands are exceptions and keep the old default of 41, because they generate the score they clamp instead of passing one through: --fasta2fastq, which has no input quality and uses this option as the value to write, and --fastq_mergepairs, which caps the computed posterior quality of a merged base.

For --fastq_mergepairs the offset in question is --fastq_ascii, not --fastq_asciiout: that command writes fastq but does not accept --fastq_asciiout, so a merged quality symbol carries the same offset the input was read with. The sum rule is stated against --fastq_ascii there, and against --fastq_asciiout everywhere else.

--fastq_qminout integer: Specify the minimum quality score used when writing fastq files. The offset (see --fastq_asciiout) plus the minimum score must be at least 33, so the value may be negative when the output offset is 64. The default is 0, which is usual for recent Sanger/Illumina 1.8+ files. Older formats may use scores between -5 and 2.

A negative minimum matters where the score passes through vsearch and may itself be negative: --fastq_convert on an offset-64 file read with a lowered --fastq_qmin writes the score back unchanged with --fastq_qminout -5, where the default 0 would raise it. It has no effect where vsearch computes the score it clamps, because such a score is derived from an error probability and is therefore never negative: --fastq_mergepairs, and --fastq_convert --fastq_solexa, which converts to the Phred scale before any output clamp applies.

--fastq_solexa: Convert quality scores from the Solexa/Illumina 1.0 scale to the Phred scale while re-encoding. A Solexa score is Q = -10 log10(p / (1 - p)), not the Phred Q = -10 log10(p), so rebasing the ASCII offset alone would leave the scores wrong (see vsearch-fastq(5)). The conversion applied is Q_phred = 10 log10(10^(Q_solexa / 10) + 1), rounded to the nearest integer; it is the identity from Solexa 10 upward, so only the fifteen lowest scores change. Requires --fastq_ascii 64, the offset the Solexa encoding uses, and implies --fastq_qmin -5, the floor of the Solexa scale, unless --fastq_qmin is given explicitly. The conversion is one-way and lossy: six Solexa pairs collapse onto a single Phred score each, and there is no option to write Solexa scores. Only --fastq_convert accepts this option.

secondary options

--bzip2_decompress: Specify that the input pipe is streaming data compressed using Huffman coding. See bzip2(1) for more details. This option is required when compressed data arrives on standard input through a pipe (‘-’), where the format cannot be detected without consuming the stream. It is not needed when reading from a regular file compressed with bzip2, nor when such a file is redirected to standard input: compression is then detected automatically, and a contradicting option is ignored (with a warning when the input is standard input). Pipes other than standard input, such as shell process substitutions and named FIFOs, are always read as uncompressed data; compressed data must arrive on standard input or as a named file.
--fastq_qmax integer: Specify the maximal quality score accepted when reading fastq sequences. Stop with an error message if a quality score higher than the specified value is read. The offset (see --fastq_ascii) plus the maximal score may not exceed 126, the last printable ASCII character: scores up to 93 with offset 33, up to 62 with offset 64. The default is the highest score the offset can represent (93 with offset 33, 62 with offset 64), so no quality score is rejected unless this option is lowered. Before version 2.32.0 the default was 41, the usual maximum for Sanger/Illumina 1.8+ files, which rejected PacBio HiFi and nanopore files outright.
--fastq_qmin integer: Specify the minimal quality score accepted when reading fastq sequences. Stop with an error message if a quality score lower than the specified value is read. The offset (see --fastq_ascii) plus the minimal score must be at least 33, the first printable ASCII character: scores down to 0 with offset 33, down to -31 with offset 64. The value may therefore be negative, which is what the negative scores of older formats require, but note that those formats are not supported (see vsearch-fastq(5)). The default is 0, which is usual for recent Sanger/Illumina 1.8+ files.
--gzip_decompress: Specify that the input pipe is streaming data compressed using Lempel-Ziv coding. See gzip(1) for more details. This option is required when compressed data arrives on standard input through a pipe (‘-’), where the format cannot be detected without consuming the stream. It is not needed when reading from a regular file compressed with gzip, nor when such a file is redirected to standard input: compression is then detected automatically, and a contradicting option is ignored (with a warning when the input is standard input). Pipes other than standard input, such as shell process substitutions and named FIFOs, are always read as uncompressed data; compressed data must arrive on standard input or as a named file.
--label_suffix string: Add the suffix string to sequence headers when writing fasta or fastq files. For example, with --label_suffix ";status=healthy", sequence header ‘>seq1’ becomes ‘>seq1;status=healthy’.
--lengthout: Add a sequence length annotation (;length=integer) to each sequence header when writing fasta or fastq files.
--log filename: Write messages to filename. Messages include program version, start and finish times, elapsed time, amount of memory available, maximum amount of memory consumed, number of cores and command line options, and if need be, command-specific informational messages, warnings, and errors. Messages are written in addition to the standard error, not instead of it; use --quiet to silence the standard error copy (see the --quiet entry for what it spares).
--no_progress: Suppress the gradually increasing progress indicator normally written to the standard error stderr(3).
--quiet: Suppress messages to the standard output stdout(3) and standard error stderr(3), except for warnings and error messages.
--relabel string: Replace sequence headers with the prefix string and a ticker (1, 2, 3, etc.). For example, with --relabel "cluster:", the first sequence header becomes ‘>cluster:1’, the second sequence header becomes ‘>cluster:2’, and so on. To retain annotations, use their corresponding options (--lengthout, --eeout, and --sizeout). Use --relabel_keep to also retain old sequence identifiers.
--relabel_keep: Retain old sequence identifiers by including them at the end of the new headers, after a space.
--relabel_md5: Replace each sequence header with the MD5 digest derived from the sequence itself. The sequence is converted to upper case, and each ‘U’ is replaced with a ‘T’ before computation of the digest. The MD5 digest is a 128-bit value (16 bytes), represented using a string of 32 ASCII characters. Each pair of characters encodes an hexadecimal value, ranging from x00 to xff. See md5(3) for more details, and --relabel_sha1 for an alternative hashing algorithm. To retain annotations, use their corresponding options (--lengthout, --eeout, and --sizeout). Use --relabel_keep to also retain old sequence identifiers.
--relabel_self: Replace each sequence header with the sequence itself. To retain annotations, use their corresponding options (--lengthout, --eeout, and --sizeout). Use --relabel_keep to also retain old sequence identifiers.
--relabel_sha1: Replace each sequence header with the SHA1 digest derived from the sequence itself. The sequence is converted to upper case, and each ‘U’ is replaced with a ‘T’ before computation of the digest. The SHA1 digest is a 160-bit value (20 bytes), represented using a string of 40 ASCII characters. Each pair of characters encodes an hexadecimal value, ranging from x00 to xff. See sha1(3) for more details, and --relabel_md5 for an alternative hashing algorithm. To retain annotations, use their corresponding options (--lengthout, --eeout, and --sizeout). Use --relabel_keep to also retain old sequence identifiers.
--sample string: Add the given sample identifier string to sequence headers when writing fasta or fastq files. For instance, if string is ‘ABC’, the text ;sample=ABC will be added to the headers. string is silently truncated at the first ‘;’ or whitespace character (space, tab, newline, carriage return, vertical tab or form feed), so such characters should not be used in string. Other characters (alphabetical, numerical and punctuations) are accepted. When nothing is left after truncation — an empty string, or one starting with ‘;’ or a blank character — vsearch issues a warning and writes a bare ;sample= annotation.
--sizein: Use the abundance annotations present in sequence headers when reading fasta or fastq file. Search for the pattern [>@;]size=integer[;]. Entries without abundance annotations are silently assumed to be of size=1.
--sizeout: Add abundance annotations to sequence headers when writing fasta or fastq files. Add the pattern ;size=integer. Existing ;size= annotations are reported unchanged; entries without one receive ;size=1. For this command --sizein is not needed: abundance annotations are always parsed from the input headers.
--xee: Strip expected error (ee) annotations from sequence headers when writing fasta or fastq files. Search for the pattern [>@;]ee=float[;]. Expected error annotations are added by the synonymous options --fastq_eeout and --eeout described in vsearch-fastx_filter(1).
--xlength: Strip sequence length annotations from sequence headers when writing fasta or fastq files. Search for the pattern [>@;]length=integer[;]. Sequence length annotations are added by the --lengthout option.
--xsize: Strip abundance annotations from sequence headers when writing fasta or fastq files. Search for the pattern [>@;]size=integer[;]. Abundance annotations are added by the --sizeout option.

ignored options

--threads positive non-null integer: Command is not multithreaded, option has no effect (a warning is printed when more than one thread is requested).

EXAMPLES

Convert a phred+64 file to the standard phred+33 encoding:

vsearch \
    --fastq_convert input.fastq \
    --fastq_ascii 64 \
    --fastq_asciiout 33 \
    --fastqout converted.fastq

Convert and clamp output quality scores to the range 0–40 (an older Illumina convention):

vsearch \
    --fastq_convert input.fastq \
    --fastq_ascii 64 \
    --fastq_asciiout 33 \
    --fastq_qmaxout 40 \
    --fastqout converted.fastq

Convert a Solexa/Illumina 1.0 file to phred+33, converting the score definition as well as the offset:

vsearch \
    --fastq_convert input.fastq \
    --fastq_ascii 64 \
    --fastq_solexa \
    --fastqout converted.fastq

CITATION

Rognes T, Flouri T, Nichols B, Quince C, Mahé F. (2016) VSEARCH: a versatile open source tool for metagenomics. PeerJ 4:e2584 doi: 10.7717/peerj.2584

REPORTING BUGS

Submit suggestions and bug-reports at https://github.com/torognes/vsearch/issues, send a pull request on https://github.com/torognes/vsearch, or compose a friendly or curmudgeont e-mail to Torbjørn Rognes (torognes@ifi.uio.no).

AVAILABILITY

Source code and binaries are available at https://github.com/torognes/vsearch.

These manual pages are also published online at https://torognes.github.io/vsearch/.

COPYRIGHT

Contact: Torbjørn Rognes torognes@ifi.uio.no, Department of Informatics, University of Oslo, PO Box 1080 Blindern, NO-0316 Oslo, Norway

This software is dual-licensed and available under a choice of one of two licenses, either under the terms of the GNU General Public License version 3 or the BSD 2-Clause License.

GNU General Public License version 3

This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

You should have received a copy of the GNU General Public License along with this program. If not, see http://www.gnu.org/licenses/.

The BSD 2-Clause License

Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:

Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.

THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

ACKNOWLEDGMENTS

We would like to thank the authors of the following projects for making their source code available:

vsearch includes code from Google’s CityHash project by Geoff Pike and Jyrki Alakuijala, providing some excellent hash functions available under a MIT license.
vsearch includes code derived from Tatusov and Lipman’s DUST program that is in the public domain.
vsearch includes public domain code written by Alexander Peslyak for the MD5 message digest algorithm.
vsearch includes public domain code written by Steve Reid and others for the SHA1 message digest algorithm.
vsearch binaries may include code from the zlib library, copyright Jean-Loup Gailly and Mark Adler.
vsearch binaries may include code from the bzip2 library, copyright Julian R. Seward.